Dataset

TREC_Clinical-Trials

The TREC Clinical Trials collections released in 2021, 2022, and 2023 are available through the TREC homepage, with corresponding papers for each year. Artur Guimarães ([email protected]) serves as the dataset curator, while inquiries regarding the original authors can be directed separately. The dataset addresses the task of linking patient profiles with appropriate clinical trials, containing entries formatted as JSON objects with fields including query identifiers, disease labels, and accompanying clinical text.

Open on Hugging Face

Data preview

Subset

Loading preview...

Dataset card

TREC Clinical Trials is a text dataset compiled from the 2021–2023 TREC Clinical Trials tracks, distributed in Parquet format on a community dataset hub. It aggregates clinical-trial-related ranking and classification examples with an associated label field.

What this dataset contains

The dataset aggregates content from the TREC 2021, 2022 and 2023 Clinical Trials tracks, which focus on information retrieval and text classification tasks over clinical-trial records. It is published under a CC-BY-SA-4.0 license and is distributed as Parquet files totaling just over 307,000 rows in the test split. The files are accessible through the pandas, Datasets and ML-Croissant libraries.

Structure and fields

Each row carries an integer id, a prompt field that itself contains a role/content pair (the user-side text used as input to a model), a completion field with its own role/content pair (the model-side response), and a string Label field used for the ranking or classification target. The prompt/completion structure suggests it was packaged for instruction-tuning or chat-style evaluation rather than as a flat table of raw documents.

How analysts use it

For investment analysts the immediate applicability is limited because the content is clinical-trial text rather than market or operating data, but it can support NLP tooling development that is later applied to pharma-relevant alternative data, such as trial-readout summarization or trial-eligibility text mining. Researchers also use the Label field as a gold standard for benchmarking retrieval and ranking systems over biomedical literature.

Documentation & references 1